Clustering is a fundamental task in biomedical research, particularly for enabling stratification of patients. Traditional clustering techniques often struggle in this domain due to the vast number of attributes to be considered and the presence of complex, non-linear data structures. Deep Clustering (DC) methods, which combine neural networks with classical clustering techniques, have emerged as a promising alternative to address this challenge. To this day, most DC research has focused on images or text, however for many biomedical problems data is in a tabular format. This work proposes and evaluates a systematic framework for assessing clustering methods on high-dimensional tabular data. The framework consists of four stages: defining the experimental setup, selecting and preprocessing datasets, implementing and executing experiments, and performing evaluation and comparative analysis. Experiments were conducted on four genomic datasets, comparing traditional clustering with and without dimensionality reduction to DC approaches (e.g. DEC, VaDE or TableDC). Results indicate classical techniques perform well in structured scenarios while DC methods show advantages in complex or imbalanced datasets. This work highlights the importance of selecting clustering techniques based on dataset characteristics. It demonstrates the potential of DC for high dimensional data.